Hugging Face Toggle Inference Endpoint State

arazzo: 1.0.1
info:
  title: Hugging Face Toggle Inference Endpoint State
  summary: Read an endpoint's current state, then pause it if running or resume it if paused.
  description: >-
    A cost-control flow over the Inference Endpoints management API. The workflow
    reads an endpoint's current status and branches on its state: when the
    endpoint is running it pauses it to stop incurring compute costs, and when the
    endpoint is paused it resumes it. Both branch actions return the updated
    endpoint record. Every step spells out its request inline so the flow can be
    read and executed without opening the underlying OpenAPI description.
  version: 1.0.0
sourceDescriptions:
- name: inferenceEndpointsApi
  url: ../openapi/hugging-face-inference-endpoints-api.yml
  type: openapi
workflows:
- workflowId: toggle-endpoint-state
  summary: Pause a running endpoint or resume a paused one based on its current state.
  description: >-
    Reads the current endpoint state and either pauses a running endpoint or
    resumes a paused endpoint.
  inputs:
    type: object
    required:
    - hfToken
    - namespace
    - endpointName
    properties:
      hfToken:
        type: string
        description: Hugging Face access token used as a Bearer credential.
      namespace:
        type: string
        description: User or organization namespace that owns the endpoint.
      endpointName:
        type: string
        description: The name of the endpoint to toggle.
  steps:
  - stepId: readState
    description: >-
      Read the endpoint's current status and branch on whether it is running or
      paused.
    operationId: getEndpoint
    parameters:
    - name: Authorization
      in: header
      value: Bearer $inputs.hfToken
    - name: namespace
      in: path
      value: $inputs.namespace
    - name: endpoint_name
      in: path
      value: $inputs.endpointName
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      currentState: $response.body#/status/state
    onSuccess:
    - name: isRunning
      type: goto
      stepId: pauseEndpoint
      criteria:
      - context: $response.body
        condition: $.status.state == 'running'
        type: jsonpath
    - name: isPaused
      type: goto
      stepId: resumeEndpoint
      criteria:
      - context: $response.body
        condition: $.status.state == 'paused'
        type: jsonpath
  - stepId: pauseEndpoint
    description: >-
      Pause the running endpoint to stop compute costs while preserving its
      configuration.
    operationId: pauseEndpoint
    parameters:
    - name: Authorization
      in: header
      value: Bearer $inputs.hfToken
    - name: namespace
      in: path
      value: $inputs.namespace
    - name: endpoint_name
      in: path
      value: $inputs.endpointName
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      newState: $response.body#/status/state
    onSuccess:
    - name: paused
      type: end
  - stepId: resumeEndpoint
    description: >-
      Resume the previously paused endpoint so it can serve inference requests
      again.
    operationId: resumeEndpoint
    parameters:
    - name: Authorization
      in: header
      value: Bearer $inputs.hfToken
    - name: namespace
      in: path
      value: $inputs.namespace
    - name: endpoint_name
      in: path
      value: $inputs.endpointName
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      newState: $response.body#/status/state
  outputs:
    previousState: $steps.readState.outputs.currentState
    pausedState: $steps.pauseEndpoint.outputs.newState
    resumedState: $steps.resumeEndpoint.outputs.newState